Since the movie Avatar was released, more than 6,000 comments have been left on IMDb and Yahoo! alone. That is more comments than there are for Star Wars (33 years since its release), and twice as many as there are for James Cameron's previous blockbuster Titanic. It has become impossible for movie viewers to keep up.
Nanocrowd developed Reaction Mapping™ technology to interpret these comments and turn them into an easy-to-use search and recommendation engine. By analyzing millions of comments, we can tell how people responded to movies, and our visitors can quickly decide which movie is right for them.
Our movie recommendations seem to be hitting the mark, because Tekzilla titled their video review of our site "Personalized movie recommendations" before we had any way for users to log in.
Think about that. We provide personalized recommendations without knowing anything about you!
Other bloggers have written about the potential of personalization based on understanding other people's reactions. AppScout wrote that our site "doesn't require human intervention like other similar sites," and TheNextWeb said "This is Web 3.0 folks."
OK, so what are Reaction Maps?
Reaction Maps are mathematical abstractions
of people's reactions to things.
A Reaction Map for a movie captures the insights and reactions contained in user commentary. Comparing Reaction Maps for different movies can answer questions like "Did people enjoy The Dark Knight?" and "What movie is most like The Dark Knight?" Reaction Maps also isolate and describe the varied reactions different people have to a single movie. Some viewers respond to "The Dark Knight" because it is an exciting movie about an action super-hero; others because it is gripping and thought-provoking. Reaction Maps understand these differences.
With the help of Reaction Maps, Nanocrowd can suggest movies you will love.
Here's how we build Reaction Maps
Step 1. We find reactions to movies.
Preferably lots of reactions. What do people think of this movie and that movie? How do they describe their favorite part? How did they feel when they watched it? Fortunately the Internet has no shortage of people's reactions to things. In fact, people seem almost possessed by a need to express their opinions. If you want to know what people thought of "The Dark Knight," there are over 15,000 reviews available for you to read on the top movie websites.
Suffice it to say, there is no shortage of reactions out there for Nanocrowd to analyze.
Step 2. Identify the cities and towns to put on our map.
We begin with linguistic and semantic analysis to understand what people mean by their comments. Read a few comments at your favorite site, and you will see why this can be a little tricky (see the examples in the box below).
What matters? What doesn't? What tells us if people liked a movie? What tells us which aspects of the movie they responded to?
Once we understand these things, we cluster reactions into logical groups we call nanogenres™ and we map them as though they were cities or towns. Each nanogenre captures a unique reaction. We name our cities with the 3 words that describe the gist of that nanogenre. They may be big bustling cities with lots of movies in them (for example, "Superhero, Exciting, Fighting" is a nanogenre that includes "The Dark Knight" and many other movies), or they may be small, unique towns with only a few movies ("Bittersweet, Comical, Confusion" including Fellini's film "I, Vitelloni").
Step 3. Place the cities and towns on the map.
Using statistical analysis, we can calculate the affinity that each nanogenre-city has to all the others cities and towns. Some cities belong together: "Superhero, Exciting, Fighting" has a bunch of suburbs, including "Comic-book, Vision, Alter-ego" and "Villain, Menace, Heroic". Other cities get placed very, very far away from these: "Broadway, Tune, Gorgeous" or "Poignant, Friendship, Aging." By figuring out how far each city and town is from every other city and town, we can create a 2-dimensional map.
Our map is already pretty interesting. What is your favorite vacation spot?
After spending some time in cities full of comic-book superheroes and villains, would you like to head south to "Gangster, Storytelling, Dramas" or go East to "Gripping, Complex, Thought-provoking?" One way leads to "Goodfellas," while the other will take you to "No Country for Old Men." Both are only a train stop away from "The Dark Knight," but they start to take you on different entertainment journeys. [In case you were wondering, yes, that is actually an outline of Australia]
Of course our real map has thousands and thousands of nanogenres, so it isn't as easy to display. Fortunately, we don't really need to show you the map, just use it to help you navigate to the right movie.
Step 4. Add the mountains and valleys.
Now that we have a map with nanogenres as cities, we know the relationship of each nanogenre to the others and we are ready to create a Reaction Map for a movie. To do that, we make our map 3-D by adding elevation to each city or town. The more a movie is like a nanogenre, the higher the mountain.
For "The Dark Knight," obviously all the nanogenre-cities that describe comic-book action movies will create a mountain range. There are some deep valleys for the nanogenres like musical comedies, and there are a few isolated mountains off in the distance like "Gangster, Storytelling, Drama."
This is a Reaction Map. Imagine the Reaction Map for several different movies, and you've got the idea. Reaction Maps are the secret sauce behind the Nanocrowd website.
Most-like movie lists
Using Reaction Maps, we overlay multiple movies and see how similar they are. We compare the Reaction Map for one movie (with all its mountains and valleys) to the Reaction Maps of others. We try to find as many overlapping mountains and valleys to find the movies that are most like each other. "Spanglish" is most like "Because I Said So" and "The Upside of Anger." "Minority Report" is most like "I, Robot" and "Blade Runner." We think our most-like lists are the best you will find, because they aren't ours at all -- they reflect the collected reactions of lots of people.
Nanogenres
Most-like lists can never really capture movies, however, because movies and people's reactions to movies are so complicated. If you tell us you want to watch a movie like "Minority Report," are you looking for an intriguing science-fiction movie like "Blade Runner" or an unpredictable thriller like "Memento?" You may enjoy "Minority Report" for both those reasons, but if you want to watch another movie like it, should we recommend "Gattaca" or "Deja vu?"
To make more accurate recommendations, we offer you 6 nanogenres that are "mountains" in the Reaction Map that describe unique reactions to that movie. Each nanogenre includes other movies that people reacted to in the same way. Now you can choose a movie that matches just exactly what you want to watch tonight.
If you liked "The Dark Knight," among the nanogenres and movies you will find "Superhero, Exciting, Fighting" and "Gangster, Storytelling, Drama" and "Gripping, Complex, Thought-Provoking"
Your evening will take a very different turn depending on which one you choose (and I bet you already know which one you would pick).
This is all very interesting (I hope it is, anyway), but how does it help you find a movie?
Reaction Mapping drives the Nanocrowd movie search engine and gives you great movie recommendations. Think back on movie suggestions you have gotten from friends. At the end of watching the movie, how often did you ask "What were they thinking?"
The problem with recommendations based on someone else's rating of a movie is that when they rate or recommend a movie, they are remembering their unique reaction. When a friend recommends "Spanglish," they may think they saw an adorable, loving, comedy-drama. They left the movie smiling and feeling uplifted. Then you go watch it and see a movie about dysfunctional relationships and marital affairs.
Everyone's reaction is unique
Everyone reacts to movies based on their mood and how they feel that moment. If you go into the theater already laughing, movies like "About Schmidt" or "The Royal Tenenbaums" will seem heartwarming or even laugh-out-loud funny. But if you are feeling discouraged, these movies will take you down a notch and you may describe them as poignant, but depressing.
Movies are complicated. People are complicated (and moody!). Not surprisingly, people's reactions to movies are complicated and unpredictable.
So, when you ask for a recommendation, how can a movie website tell you what movie you are going to enjoy? How can the website know which parts of which movies you might like, and also know what mood you are in right now? Reaction Mapping makes it all possible.
Matching your interest and mood
By displaying and letting you choose nanogenres, we help you pick a movie that matches your mood. If you are feeling giddy, you will want to watch different things than if you are feeling distraught. You may remember how much you enjoyed "The Whole Nine Yards," but depending on your mood you may want to avoid the movies in the nanogenre "Hit-man, Quirky, Cold-blooded" and prefer "Humor, Unexpected, Delight" instead. Each nanogenre contains movies that have something in common with "The Whole Nine Yards," but they are from very different mountains in its Reaction Map.
At their core, Reaction Maps are complex mathematical abstractions of people's reactions to movies. People's reactions are unpredictable and inconsistent. The commentary they write is all over the map. Building a Reaction Map that can describe how people respond to something as complex as a movie would seem an impossible task, but visit our site and see the results for yourself. Try it out. I think you will enjoy how quickly we get inside your head to recommend movies.
posted by Roderic March
Posts
